MediumNLPnlp-094

chrF-lite Character F-score

Problem

Return the character unigram F1 between candidate and reference strings. Count clipped per-character matches, then F1 = 2PR/(P+R) with precision = matches / len(candidate) and recall = matches / len(reference).

Return 0.0 when both strings are empty or there are no matches.

Hint

Characters are matched with clipping, exactly like a bag-of-words F-score.

Test cases

case 1

in: ["abc","abc"]

expected: 1

case 2

in: ["abc","abd"]

expected: 0.6666666666666666

case 3

in: ["ab","a"]

expected: 0.6666666666666666

case 4

in: ["abc","xyz"]

expected: 0

case 5

in: ["aab","ab"]

expected: 0.8

Starter codePython

def chrf_lite(candidate, reference):
    # Your code here
    pass

5 test cases · run your code against them in the editor